Operational police horses must perform under challenging circumstances, so the selection of suitable equine training candidates should include an evaluation of their personalities. We describe the developmental phase of a project aimed at customizing selection tests for mounts at the Belgian Federal Police. The experiment evaluated the influence of preceding activity and multiple testing on behavior during personality tests. Using a crossover design, three different preceding activities were compared: straight out of their box, after 30 minutes tied outside alone, and after 30 minutes alone in a paddock. A sample of 30 active police horses was tested, in three equal groups with a different treatment order per group. The personality tests included an arena and a sudden object test, measuring behavioral responses. For the sudden object test, we also recorded distance from a descending umbrella. Likert scores of 1 to 5 were used to score the ease of handling when leading the horse into and out of the arena. Order and treatment effects were analyzed with mixed models. Treatment effects were found in the arena test for 10 of 17 variables, including all measures of stand, canter and roll, and buck/jump rate. Order effects were present for stand rate in the arena and sudden object test, for rate and mean bout length of walk in the sudden object test, and for the scores of release at entering the arena and approach at the end of the testing. After all horses had been tested individually, 10 horses from the current sample were tested as duos, after coming straight out of their boxes (stables) to evaluate the effect of solitary testing and the influence of equine company as it prevails in the workplace. A Wilcoxon signed rank test that compared the duo results to those of the individual tests for the same treatment, showed a significant difference in only two of 52 variables (stand mean bout length in the arena test and Likert score for entering the arena and stopping). Conversely, only five variables had significant ...